How to Audit Your AI Search Ranking in 20 Minutes

Running an AI search test in your own ChatGPT or Claude account is not an audit. It is a false positive. Your personal AI instance has memory of who you are. It hedges toward showing you results where you appear. Someone else searching the same query on a different device may see a completely different set of citations, with no mention of your company at all.

An accurate audit requires fresh LLM sessions with no memory, run at scale across the searches your actual buyers are typing. That is what this workshop demonstrates, live, on a real company.

Catalyst has worked with 150 companies over the last 2 years to help them rank consistently in AI search, from pre-revenue startups to companies doing $100M in revenue. The patterns that drive and block AI citations are the same across all of them.

Why Does Testing Your Own AI Account Give You the Wrong Answer?

Every major LLM maintains session context and, in some cases, persistent memory. When you ask "best real estate agents in Southern California" in your own account, the model knows your identity and biases the output accordingly.

A clean audit does the opposite. It opens fresh sessions with no prior context, runs 100 queries across the searches your buyers are realistically typing, and aggregates results across 5 platforms: ChatGPT, Claude, Gemini, Perplexity, and Google AI Overview. The result is a spectrum of where you appear, how often, and what the model says about you when it has no reason to favor you.

The company audited live in this workshop was showing up at 77% on most LLMs but barely registering on Google AI Overview. That gap has a specific cause and a specific fix.

What Does the AEO Score Actually Measure?

The audit scores every company across 3 pillars that map directly to how LLMs decide whether to cite a source:

| Pillar | What It Measures | Common Gaps | |---|---|---| | Credible | Off-site citations from high-authority, industry-specific sources | Domain authority under 30, thin press coverage, no Wikipedia presence | | Readable | Technical setup: schema markup, site speed, structured data, LLM.txt | Missing FAQ schema on subpages, no llms.txt, slow load times | | Substantive | On-site content depth and originality | Generic copy with no original data or expert opinion, no how-to guides |

The company in this workshop scored 56 overall. That is above average but still left significant citation frequency on the table, especially on Google AI Overview, which scores separately from the other LLMs and follows different ranking rules. For a deeper breakdown of how each LLM weights different content signals, see How B2B Marketing Teams Get Named in AI Search.

Why Do Different LLMs Rank You Differently?

Not all LLMs pull from the same sources, so the same company can rank consistently on Perplexity and be nearly invisible on Google AI Overview. The model behavior differences matter:

ChatGPT was trained heavily on Wikipedia and structured web content. It responds best to formal, wiki-style information architecture. Perplexity, Claude, and Grok weight third-party citations and community platforms more heavily: Reddit, G2, LinkedIn, and industry publications. Google AI Overview is an extension of Google Search's E-E-A-T standards, which means it prefers pages already ranking in organic search and penalizes content that lacks demonstrated expertise and real-world experience.

The implication is that the fastest path to Google AIO visibility runs through SEO fundamentals: domain authority, structured content, and original expert perspective embedded in the page itself, not listed as credentials in a bio.

What Are the 3 Biggest Gaps Holding Companies Back From AI Citations?

The live audit surfaced 3 gaps that appear consistently across the 150 companies Catalyst has worked with. Each one is fixable, and each can shift citation frequency within days of a change.

Gap 1: Domain authority under 30 with thin press coverage

LLMs use domain authority as a credibility proxy. A domain authority of 11, like the company audited in this workshop, tells the model that third-party sources have not validated the site at scale. Wikipedia presence and press coverage are the fastest paths to closing this gap because they directly feed the source graph LLMs draw citation decisions from. The fastest path to citation parity is press coverage and a Wikipedia entry, not more on-site content.

Gap 2: FAQ schema missing from internal pages

FAQ schema on a homepage is not enough. Each subpage covering a specific topic, market, or service needs its own FAQ block, structured so the LLM can lift the question-and-answer pair verbatim. The company audited here had FAQs on the homepage but not on any location-specific subpages. A buyer searching "best real estate agent in Newport Beach" triggered queries that landed on the Newport Beach page and found no structured answer. The model skipped the citation. How to Get ChatGPT, Claude, and Gemini to Recommend Your B2B Company covers how to structure this for service businesses.

Gap 3: No LinkedIn entity authority

When an LLM wants to verify a company's credibility, it cross-references third-party sources: the company's LinkedIn page, the founder's profile, G2, industry forums, and brokerage or platform listings. A company with no active LinkedIn presence gives the model fewer anchor points to construct an entity understanding from. Will flagged LinkedIn as one of the fastest-moving authority signals he has seen shift in the last 6 months. Posting market data, expert commentary, and original perspective 2 to 3 times a week builds entity authority over time and directly improves citation frequency on Claude, Perplexity, and Grok.

How Do You Run This Audit on Your Own Company?

The process from the workshop is repeatable on any company in any industry:

1. Run fresh LLM queries (not in your personal account) across all 5 platforms using buyer language, not your brand name. Think: what would a buyer type to find a solution to the problem you solve?
2. Check your score across the 3 pillars: Credible, Readable, Substantive.
3. Identify which LLMs you appear in versus which you do not. Each gap points to a different fix.
4. Download the AI Memory Pack from the audit results — a markdown file with all findings you can paste into any LLM to generate a prioritized action plan.
5. Re-run weekly. AI search results can shift within days of a content change. Monthly reviews miss the signal.

The audit used in this workshop runs 100 fresh queries across 5 LLMs and produces a scored report with specific recommendations. For context on how to act on the authority gap specifically, see How to Win AI Search Citations Against Higher-Authority Competitors.

The Takeaway

The companies winning AI search right now are publishing content that LLMs can lift verbatim: structured FAQs on every relevant subpage, data-backed content with original expert opinion layered in, and consistent third-party validation from LinkedIn, press coverage, and industry platforms. The audit takes 20 minutes. Making a change, whether adding FAQ schema, posting to LinkedIn, or building a market data page, can move citation frequency within days.

The one action to take this week: run the audit in a fresh session (not your own AI account), and identify whether your gap is in Credibility, Readability, or Substance. That single answer tells you exactly where to spend the next 30 days.